AI Data Lead (Generative AI, LLM)

3 Lý do để gia nhập công ty

• AI-native culture with enterprise AI tools
• Flexible hybrid work, focused on outcomes
• Small team, high ownership

Mô tả công việc

WHAT WE ARE

Zeya Labs is a small AI-native engineering team based in Ho Chi Minh City. We are the central technology delivery engine for a diversified group spanning financial services, F&B, retail, hospitality, and real estate.

We build shared platform utilities, run AI pilots across group entities, and own the product thinking that makes those things useful rather than merely functional. We are not a consultancy. We do not hand off specifications. We build, run, and own what we ship.

We are early. The org is being stood up now. If you need things to be settled before you can do your best work, this probably isn't the right move.

WHAT THIS ROLE IS

Every group entity generates data on its own terms — different systems, different structures, no shared platform. It's tempting to respond to that with a target-state architecture: map every source, design the ideal schema, build the platform that fixes it once and for all. We are explicitly not doing that. Nobody is commissioning a data platform build here.

What actually happens is narrower and more useful: a shared utility or targeted solution gets blocked because the data behind it is messy, fragmented, or locked in a format nobody's cleaned up. This role exists to unblock that — using AI to do in days what used to take a data engineering team a quarter.

That's the mindset shift this role needs. Not "what's the correct long-term data model," but "what's the smallest, smartest thing I can point at this mess to make it usable, right now, for the thing that's actually blocked." Sometimes that's an AI agent doing extraction and reconciliation across messy source systems. Sometimes it's a narrow pipeline.

Concretely, this role owns three things:

• Finding the highest-leverage problem, not the biggest one. Not surveying every system the group has. Finding the specific data problem that's actually blocking an AI pilot or shared utility right now, and going after that — not the tidiest or most complete problem, the one that unblocks something real.

Using AI to solve the data problem, not just engineering around it. Reaching for AI-native approaches first — agents, LLM-based extraction, matching, and reconciliation — instead of defaulting to hand-built pipelines and manual schema work. This is the advanced-thinking bar: staying current on what AI can now do to a data mess that used to require heavy traditional engineering, and knowing when it genuinely can't and something more conventional is needed.

• Building toward specific goals. Every fix ties to a specific goal — an AI pilot that's stuck, a utility that needs clean input. No platform roadmap, no multi-quarter build. Ship the narrow thing, move to the next blocker.

Yêu cầu công việc

WHO WE'RE LOOKING FOR

You think about data problems the way an AI-native builder does. Given a mess, your first instinct is "can I point an AI agent at 80% of this and get something usable in a week," not "what's the correct long-term schema." This is a hard filter: someone whose reflex is comprehensive design is the wrong hire for this role, however good the design would be.

You need to be genuinely advanced in your thinking here — not a traditional data engineer who's picked up ChatGPT as a productivity tool, but someone whose actual toolkit for solving data problems already includes AI-native techniques as the default, not the add-on. You should be able to point to real messes you've solved this way, not just pipelines you've hand-built.

You're allergic to boiling the ocean. You can tell the difference between a data problem that's genuinely hard and one that just looks hard because nobody's pointed the right AI tool at it yet — and you'd rather ship the narrow fix than spend a month justifying the comprehensive one.

You're comfortable across a multi-sector portfolio — financial services, F&B, retail, hospitality, real estate — and you don't need every source system mapped before you'll touch the one that's actually in your way.

A few things that matter to us:

• You default to the smallest fix that unblocks the next real initiative, not the most complete one.
• You'd rather point an AI agent at a data mess and get most of the way there in days than spend weeks designing the perfect model.
• You're openly skeptical of comprehensive data governance or platform programs that take a year and unblock nothing in the meantime.
• You can hold that position with senior stakeholders — including when someone wants the big design instead of the narrow fix — without turning it into a confrontation.

Tại sao bạn sẽ yêu thích làm việc tại đây

.